Near-infrared gallium nitride two-dimensional photonic crystal platform on siliconI. Roland, Y. Zeng, Z. Han, X. Checoury, C. Blin, M. El Kurdi, A. Ghrib, S. Sauvage, B. Gayral, C. Brimont, T.

Guillet, F. Semond, and P. Boucaud

Citation: Applied Physics Letters 105, 011104 (2014); doi: 10.1063/1.4887065 View online: http://dx.doi.org/10.1063/1.4887065 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/105/1?ver=pdfcov Published by the AIP Publishing

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 193.48.18.45

On: Mon, 07 Jul 2014 13:09:25

Near-infrared gallium nitride two-dimensional photonic crystal platformon silicon

I. Roland,1 Y. Zeng,1 Z. Han,1 X. Checoury,1 C. Blin,1 M. El Kurdi,1 A. Ghrib,1 S. Sauvage,1

B. Gayral,2,3 C. Brimont,4 T. Guillet,4 F. Semond,5 and P. Boucaud1,a)

1Institut d’Electronique Fondamentale, CNRS - Univ. Paris Sud 11, Batiment 220, F-91405 Orsay, France2Univ. Grenoble Alpes, INAC-SP2M, CEA-CNRS group “Nanophysique et Semiconducteurs,”F-38000 Grenoble, France3CEA, INAC-SP2M, CEA-CNRS group “Nanophysique et Semiconducteurs,” F-38000 Grenoble, France4Universit�e Montpellier 2, Laboratoire Charles Coulomb UMR 5221, F-34905 Montpellier, France5CRHEA-CNRS, Rue Bernard Gr�egory, F-06560 Valbonne, France

(Received 9 June 2014; accepted 24 June 2014; published online 7 July 2014)

We demonstrate a two-dimensional free-standing gallium nitride photonic crystal platform

operating around 1550 nm and fabricated on a silicon substrate. Width-modulated waveguide

cavities are integrated and exhibit loaded quality factors up to 34 000 at 1575 nm. We show the

resonance tunability by varying the ratio of air hole radius to periodicity, and cavity hole

displacement. We deduce a �7.9 dB/cm linear absorption loss for the suspended nitride structure

from the power dependence of the cavity in-plane transmission. VC 2014 AIP Publishing LLC.

[http://dx.doi.org/10.1063/1.4887065]

The nitride semiconductor materials are considered as

the most appropriate materials for short wavelength and

ultra-violet nanophotonics.1–3 More recently, there has been

an increased interest for using nitrides like gallium nitride or

aluminum nitride in the near-infrared at telecom wave-

lengths. As opposed to silicon which is centrosymmetric, the

nitride semiconductors exhibit a non-vanishing second-order

nonlinear susceptibility4 which can be used for second-order

nonlinear processes (harmonic generation, frequency conver-

sion). The mechanical properties and piezoelectricity of the

nitrides could also be exploited for optomechanics.5 As com-

pared to III–V materials like InP or GaAs or silicon materi-

als, the third order susceptibility for two-photon absorption

is strongly decreased since two-photon absorption at 1.55 lm

falls in the transparency window of gallium nitride.6 The

large band gap of the nitride is thus a very important asset.

Furthermore, the nitrides exhibit a high thermal conductivity

and could possibly support much higher pump power den-

sities as compared to silicon. These properties are all the

more interesting that the nitride can be integrated on silicon,

thus offering a new platform for silicon photonics, using

nitride as passive or active materials.7,8 Integrating photonic

crystals onto this platform brings higher functionalities for

light propagation management and the realization of com-

plex routing architectures with couplers, delay lines, etc. The

index contrast between nitride (�2.3 at telecommunication

wavelength) and air is important, thus allowing to fabricate

cavities with high Q factors and small modal volumes.9

One of the interests associated with near-infrared nitride

photonics platform is the possibility to develop two-

dimensional planar photonic circuits where light can be

coupled from the cleaved facets. This type of two-

dimensional platform offers higher flexibility for characteri-

zation and future integration on chip. A first demonstration

of such gallium nitride platform on silicon has been reported

recently.10 The investigated structures were composed of

suspended ridge-access waveguides and two-dimensional

photonic crystals. Quality factors up to 5400 for double het-

erostructure cavities at 1611 nm with in-line coupling were

demonstrated with this platform.11 It is worth noting that

very high quality factors up to 148 000 at 1530 nm were also

obtained with suspended aluminum nitride nanobeam cav-

ities fabricated with sputter-deposited AlN-on-insulator sub-

strates.12 In the latter case, input and output light were

distributed with grating couplers. One of the drawback is

that the tuning of the nanobeam cavity resonance is less

practical as compared to a two-dimensional photonic crystal

where heating elements and metal pads can be deposited in

close proximity with the photonic crystal cavities.13

In this work, we demonstrate a gallium nitride two-

dimensional photonic crystal platform operating around

1550 nm. The platform consists of free-standing ridge nitride

waveguides suspended by nanotethers and suspended two-

dimensional photonic crystal nitride membranes grown on a

silicon substrate. We report loaded quality factors for width-

modulated waveguide cavities up to 34 000. We illustrate the

tuning capacity of this system by varying the photonic crys-

tal parameters. We have finally investigated the cavity

response as a function of the incident power. We deduce the

value of the residual linear absorption in suspended gallium

nitride from the red-shift of the resonance at high pump

powers.

The studied samples were grown by molecular beam

epitaxy on a Si (111) substrate. The nitride layer consists of

a thin AlN layer, around 45 nm thick, epitaxially grown on

silicon followed by a 300 nm thick GaN layer. Two samples

with total thicknesses of 330 and 345 nm were studied. There

is some residual stress in the GaN layer as well as a strain

gradient along the vertical direction. This strain gradient

results from the compressive strain relaxation when growing

GaN on AlN. The relaxation profile in GaN depends on the

a)Electronic mail: philippe.boucaud@ief.u-psud.fr. URL: http://pages.ief.

u-psud.fr/QDgroup

0003-6951/2014/105(1)/011104/4/$30.00 VC 2014 AIP Publishing LLC105, 011104-1

APPLIED PHYSICS LETTERS 105, 011104 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 193.48.18.45

On: Mon, 07 Jul 2014 13:09:25

structural quality of the thin AlN buffer layer. The residual

stress results from this compressive strain gradient and the

tensile strain induced by the difference of thermal dilatation

coefficients between nitrides and the silicon substrate when

going from epitaxy temperature to room temperature. The

structural quality of the thin AlN layer has an impact on the

residual stress and on the structural quality of the GaN layer.

For processing, an oxide layer was first deposited on the

sample. The electronic lithography of the waveguides and

photonic crystal patterns was then performed in a single step.

The pattern was transferred by reactive ion etching into the

oxide layer. We then did etch the nitride material using

chlorine-based inductively coupled plasma (ICP) etching

with a mixture of Cl2, BCl3, and Ar gases (25/10/5 SCCM).

The ICP etching rate is 400 nm/min. The selectivity with the

oxide mask is 1:16. The oxide mask was removed after ICP

etching using a buffered fluorhydric acid solution and before

the release of the membrane. Dies with a total length of

500 lm were obtained by saw dicing. The silicon was then

selectively etched using XeF2 to release the suspended struc-

tures, with an etching depth between 1.5 and 5 lm. The two-

dimensional photonic crystal has a length between 18 and

24 lm. Light is guided by a W0.98 waveguide corresponding

to a width of 0:98ffiffiffi3p

a, where a is the photonic crystal lattice

periodicity (580 nm). The hole radius to periodicity ratio was

varied around 0.23. We have designed width-modulated

waveguide cavities following the A1 design presented in

Ref. 14 (see inset of Fig. 3(b)). The design of these cavities

consists in shifting laterally 30 holes around the cavity center

with a maximum hole displacement of 15 nm. The cavity is

coupled to the access ridge waveguides by barrier coupling

waveguides with a length that can be varied between 13 and

17 lattice periodicities. These in-line coupling waveguides

control the cavity transmission and have an impact on the

loaded quality factor. Light is injected in the photonic crystal

from the diced sample edge through ridge waveguides sus-

pended by nanotethers following the design presented in

Ref. 15. The tethers have a 140 nm width at the waveguide

side and 350 nm width at the blanket side. The number of

supporting tethers was varied between 16 and 36 for the

whole suspended structure corresponding to a spacing of 12

and 27 lm. We used a non-periodical tether positioning in

order to avoid Bragg resonances with large amplitudes in the

transmission spectra. For transmission measurements, light

is injected with lensed fibers through tapers in order to opti-

mize the coupling efficiency. The taper length is 12 lm.

Figure 1(a) shows an optical microscopy image of the whole

structure. Figure 1(b) shows a scanning electron microscopy

image of the transition between the ridge waveguide and the

photonic crystal and Figure 1(c) shows a zoom around the

access taper. The coupling efficiency was first characterized

by using suspended waveguides without photonic crystals.

Fig. 1(d) shows the transmission losses as a function of the

number of tethers. The transmission in dB depends linearly

on the number of tethers. The extrapolation of the curve to

zero tether indicates that the coupling losses are 10 dB, if we

neglect the linear absorption losses, as justified below. If we

subtract the 3 dB losses measured without sample, we obtain

a total 7 dB loss corresponding to coupling losses of 3.5 dB

for each taper. These values could obviously be optimized

by design and fabrication but we are already very close to

the performances obtained with similar structures on silicon.

For the structure with 16 tethers, the scattering losses are

5.5 dB for a 500 lm length, corresponding to a 11 dB/mm

propagation loss for this gallium nitride suspended structure.

This value is equivalent than the one reported in Ref. 10

(12 dB/mm).

Figure 2 shows a transmission spectrum measured for a

structure with a 24 lm photonic crystal and supported by 14

nanotethers on each side of the photonic crystal. A width-

modulated waveguide cavity with a maximum hole displace-

ment of 12 nm is located at the center of the photonic crystal.

FIG. 1. (a) Optical, microscopy image of the two-dimensional photonic

crystal platform with the photonic crystal at the center, and the access wave-

guides suspended by nanotethers. The total structure length is 500 lm. (b)

Zoom at the transition between the ridge waveguide and the photonic crys-

tal. (c) Zoom around the coupling taper for light injection/collection with

lensed fibers. (d) Transmission losses for 500 lm long structures without

photonic crystals as a function of the number of tethers. The full line is a lin-

ear fit. Its extrapolation to zero gives the coupling losses due to both tapers

and experimental setup.

FIG. 2. (a) Transmission of a structure constituted of a 24 lm long photonic

crystal with a width-modulated waveguide cavity. The number of supporting

tethers on each side of the photonic crystal is 14. The maximum displace-

ment of the cavity hole is 12 nm. The barrier coupling length between cavity

and ridge waveguide is 17 a. The inset shows a zoom around the resonance

of the cavity fundamental mode. A loaded quality factor of 34 000 is meas-

ured. (b) Comparison between the experimental transmission spectrum and

the calculated one (red curve) using Meep software. The Q factor of the res-

onance is limited by the modeling time.

011104-2 Roland et al. Appl. Phys. Lett. 105, 011104 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 193.48.18.45

On: Mon, 07 Jul 2014 13:09:25

The lattice periodicity a is 580 nm and the air hole radius ra-

tio to periodicity ra is 0.23. The length of the coupling wave-

guides between the photonic crystal edge and the cavity is

17 a. The thickness of the AlNþGaN layer is 345 nm. In

order to obtain the full dynamical range, the spectrum is

measured with two sensitivities on the oscilloscope. Outside

the photonic crystal band gap, the transmission losses are

�23 dB at 1520 nm, including 19.5 dB losses associated with

tapers and tethers. The photonic band gap of the W0.98

waveguide is clearly observed between 1570 and 1590 nm.

The oscillations observed around the base line at �52 dB are

attributed to a residue due to a TM-polarized mode. The visi-

bility of this residue depends on the silicon etching depth

and on the taper size. Two factors can explain the conversion

of TE-polarized light to TM polarized light. The vertical

stack is intrinsically asymmetric as aluminum nitride and

gallium nitride have different refractive indices. The etching

of the holes is not perfectly vertical, with a 4� residual incli-

nation, thus leading to TE-to-TM conversion.8 The striking

feature reported in Fig. 2 is the presence of a resonant mode

around 1575 nm which is attributed to the fundamental mode

of the width-modulated waveguide cavity. The inset of

Fig. 2 shows a zoom around this resonance, with a measured

loaded quality factor of 34 000, much larger than those previ-

ously reported in similar architectures.11 The loaded quality

factor Ql results from the cavity quality factor Qcav and the

coupling quality factor Qc following 1Ql¼ 1

Qcavþ 1

Qc. The cav-

ity transmission at resonance is equal to Ql

Qc

� �2

. Considering

a �26.5 dB transmission at resonance between the access

waveguides before and after the photonic crystal, it gives a

coupling quality factor of 718 500 and a cavity quality factor

of 35 690. This value still remains significantly smaller than

the theoretical intrinsic value (490 000). This origin of the

difference is discussed below. Fig. 2(b) shows a comparison

between the transmission spectrum and the theoretical one as

obtained by finite difference in time domain modeling using

the Meep software.16 A good agreement is obtained, in par-

ticular, for the band gap spectral width and the resonant posi-

tion of the fundamental mode of the width-modulated

waveguide cavity. The modeling does not account for the re-

sidual TM-mode.

Figure 3 shows examples of the large tunability that can

be achieved with this system and the capability to tune the

resonance wavelength into the C-band (1530–1565 nm). The

measurements were performed with the 330 nm thick sample

and the residual TM mode is hardly visible in these measure-

ments. Fig. 3(a) shows the transmission dependence as a

function of the ra ratio. The photonic crystal band gap

increases significantly as ra increases. As expected, the cavity

transmission also decreases as ra increases while the length of

the photonic crystal barrier waveguide is kept constant.

Fig. 3(b) depicts the transmission dependence as a function

of the hole displacement of the width-modulated waveguide

cavity. The increase of displacement leads to a red-shift of

the resonance cavity, along with an increase of the quality

factor. We have also checked the dependence of the cavity

transmission amplitude as a function of the length of the cou-

pling barrier waveguide (not shown). Decreasing this param-

eter leads to an increase of the transmission at the expense of

the loaded quality factor.

Two-photon absorption is a strongly limiting factor of

Si, InP, or GaAs for photonic circuit applications around

1.55 lm, as it leads to saturation of transmission in resonant

cavities as pump power is increased. This saturation results

from resonance shift due to a combination of free carrier

absorption-dispersion and thermal effect. Due to its much

larger band gap, it is expected that residual two-photon

absorption is much weaker in GaN. We have thus investi-

gated the pump power dependence of a cavity similar to the

one presented in Fig. 2 (Ql factor around 31 000, number of

tethers 16). Figure 4 shows the measurements at different

FIG. 3. (a) Transmission dependence as a function of ra ratio. The maximum

hole displacement for the cavity is 9 nm. The barrier coupling length is 13 a.

The sample thickness is 330 nm. (b) Transmission dependence as a function

of lateral displacement for the width-modulated waveguide cavity. The inset

shows a schematics of the lateral hole displacements for the cavity. The blue

and yellow arrows correspond to 2/3 and 1/3 of the main displacement.

FIG. 4. Transmission of the resonance associated with a width-modulated

waveguide cavity as a function of the pump power measured at the laser out-

put. The Ql factor at low excitation is 31 000. The number of supporting

tethers is 16.

011104-3 Roland et al. Appl. Phys. Lett. 105, 011104 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 193.48.18.45

On: Mon, 07 Jul 2014 13:09:25

incident pump powers measured at the laser output. The res-

onance is red-shifted as the pump power is increased. We at-

tribute this red-shift to a thermal effect due to residual linear

absorption and we neglect the dispersion associated with free

carrier generation. As the power continues to increase, the

asymmetry of the resonance becomes more pronounced as a

consequence of the wavelength sweep from short to long

wavelength.17 The output power follows a linear dependence

as a function of the input power, indicating that the losses

stem from linear absorption. One can estimate the absorbed

power Pabs due to residual linear absorption from the spectral

shift of the resonance following dk ¼ @n@T RthPabs

kn and

Pabs ¼ 2a c c Ql

ffiffiffiffiffiffiffiffiTrmax

p

n x0Pin ¼ 2c

Q2l

QcQabsPin, where Qabs ¼ 2pn

ak and c

is the fraction of the mode energy in the material (86%), n is

the refractive index, @n@T is the thermo-optic coefficient, Rth is

the thermal resistance. Pin is the input power at the entrance

of the photonic crystal. a is the unknown linear absorption,

Trmax is the cavity transmission at resonance and low power.

The contribution of thermal dilation to the wavelength shift

is negligible. By considering a thermo-optic coefficient @n@T

for GaN (Ref. 18) of 10�4 K�1 and a thermal resistance Rth

of 5780 (K/W),19 one deduces a linear absorption a of

1.8 6 0.4 cm�1, i.e., 7.9 dB/cm at 1575 nm. We can note

that a 7.9 dB/cm propagation loss gives an absorption

quality factor Qabs of about 49 800. The measured quality

factor is thus mainly dominated by the linear absorption

loss. The cavity quality factor Qcav is equal to 32 200

1Ql¼ 1

Qcavþ 1

Qc; Qc ¼ 824 824

� �and the intrinsic quality

factor Qint as given by 1Qcav¼ 1

Qintþ 1

Qabsis equal to 91 000. We

can also observe that the value of linear absorption indicates

that the losses as discussed in Fig. 1(d) are dominated by the

scattering losses due to the suspended nanotethers, thus mak-

ing the measurement of the absorption losses challenging.

The sensitivity of the photonic crystal resonance allows one

to estimate this low absorption value.

In conclusion, we have demonstrated a free standing

two-dimensional gallium nitride photonic crystal platform

operating at the telecommunication wavelength. We have

fabricated width-modulated waveguide cavities with quality

factors up to 34 000, significantly larger than those previ-

ously published. We have investigated the pump power de-

pendence of the cavity transmission. We have deduced a

linear absorption of 7.9 dB/cm for the free-standing nitride,

an absorption difficult to estimate in standard propagation

measurements due to the predominance of scattering by

nanotethers. This two-dimensional platform appears as very

promising for nonlinear experiments with nitride

semiconductors.

This work was supported by Agence Nationale de la

Recherche under QUANONIC convention (ANR 2013BS10-

0010-03). This work was also partly supported by the

RENATECH network. We acknowledge support from

GANEX (Grant No. ANR-11-LABX-0014). GANEX

belongs to the public funded “Investissements d’Avenir”

program managed by the French National Research Agency.

1D. Sam-Giao, D. N�eel, S. Sergent, B. Gayral, M. J. Rashid, F. Semond,

J. Y. Duboz, M. Mexis, T. Guillet, C. Brimont, S. David, X. Checoury,

and P. Boucaud, Appl. Phys. Lett. 100, 191104 (2012).2S. Sergent, M. Arita, S. Kako, K. Tanabe, S. Iwamoto, and Y. Arakawa,

Appl. Phys. Lett. 101, 101106 (2012).3M. Stegmaier, J. Ebert, J. M. Meckbach, K. Ilin, M. Siegel, and W. H. P.

Pernice, Appl. Phys. Lett. 104, 091108 (2014).4Y. Fujii, S. Yoshida, S. Misawa, S. Maekawa, and T. Sakudo, Appl. Phys.

Lett. 31, 815–816 (1977).5C. Xiong, W. H. P. Pernice, X. Sun, C. Schuck, K. Y. Fong, and H. X.

Tang, New J. Phys. 14, 095014 (2012).6C.-K. Sun, J.-C. Liang, J.-C. Wang, F.-J. Kao, S. Keller, M. P. Mack, U.

Mishra, and S. P. DenBaars, Appl. Phys. Lett. 76, 439–441 (2000).7C. Xiong, W. H. P. Pernice, and H. X. Tang, Nano Lett. 12, 3562–3568

(2012).8D. N�eel, I. Roland, X. Checoury, M. El Kurdi, S. Sauvage, C. Brimont, T.

Guillet, B. Gayral, F. Semond, and P. Boucaud, Adv. Nat. Sci.: Nanosci.

Nanotechnol. 5, 023001 (2014).9X. Checoury, D. N�eel, P. Boucaud, C. Gesset, H. Girard, S. Saada, and P.

Bergonzo, Appl. Phys. Lett. 101, 171115 (2012).10U. Dharanipathy, N. Vico Trivino, C. Yan, Z. Diao, J.-F. Carlin, N.

Grandjean, and R. Houdre, Opt. Lett. 37, 4588–4590 (2012).11N. Vico Trivino, U. Dharanipathy, J.-F. Carlin, Z. Diao, R. Houdre, and N.

Grandjean, Appl. Phys. Lett. 102, 081120 (2013).12W. H. P. Pernice, C. Xiong, C. Schuck, and H. X. Tang, Appl. Phys. Lett.

100, 091105 (2012).13L.-D. Haret, X. Checoury, F. Bayle, N. Cazier, P. Boucaud, S. Combri�e,

and A. de Rossi, Opt. Express 21, 10324–10334 (2013).14E. Kuramochi, M. Notomi, S. Mitsugi, A. Shinya, T. Tanabe, and T.

Watanabe, Appl. Phys. Lett. 88, 041112 (2006).15Z. Han, X. Checoury, D. N�eel, S. David, M. El Kurdi, and P. Boucaud,

Opt. Commun. 283, 4387–4391 (2010).16A. F. Oskooi, D. Roundy, M. Ibanescu, P. Bermel, J. Joannopoulos, and S.

G. Johnson, Comput. Phys. Commun. 181, 687–702 (2010).17T. Uesugi, B. S. Song, T. Asano, and S. Noda, Opt. Express 14, 377–386

(2006).18N. Watanabe, T. Kimoto, and J. Suda, J. Appl. Phys. 104, 106101 (2008).19L.-D. Haret, A. Ghrib, X. Checoury, N. Cazier, Z. Han, M. El Kurdi, S.

Sauvage, and P. Boucaud, Opt. Lett. 39, 458–461 (2014).

011104-4 Roland et al. Appl. Phys. Lett. 105, 011104 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP: 193.48.18.45

On: Mon, 07 Jul 2014 13:09:25